Wolfram LLM Benchmarking Project: leaderboard
Ongoing Wolfram Language code-generation benchmark from Wolfram Research. Models produce executable Wolfram Language code and are scored on syntax validity and functional correctness.
Metric: Correct Functionality (%). Source: www.wolfram.com. Status: years away from saturation. 535 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 (Thinking) | 72.5 |
| 2 | Gemini 3 Pro (Preview) | 71.2 |
| 3 | Gemini 3.1 Pro (Preview) | 71.1 |
| 4 | Gemini 3 Flash (Preview) | 69.2 |
| 5 | Claude Opus 4.5 (Thinking) | 68.9 |
| 6 | GPT-5.5 (xHigh) | 68.8 |
| 7 | Qwen 3.6 Max Preview | 68.7 |
| 8 | Claude Opus 4.7 | 68.4 |
| 9 | GPT-5.6 Sol (xHigh) | 68.3 |
| 10 | Muse Spark 1.1 | 67.8 |
| 11 | GLM-5.1 (Thinking) | 67.8 |
| 12 | Qwen 3.7 Max (2026-06-08) | 67.7 |
| 13 | GPT-5.6 Sol (Medium) | 67.4 |
| 14 | Kimi K2.6 (Thinking) | 67.3 |
| 15 | Grok 4.20 (Reasoning) | 66.3 |
Interactive version: theaggregate.ai/benchmark?slug=wolfram-llm-benchmarking-project · How It Works · Data refreshed daily, snapshot 2026-09-05.